Fast Descent Methods for LPs With No Matrix Inversions

Authors

  • Katta G. Murty Department of IOE, University of Michigan, Ann Arbor, MI 48109-2117, USA

Abstract

Existing software implementations for solving Linear Programming (LP) models are all based on full matrix inversion operations involving every constraint in the model in every step. This linear algebra component in these systems makes it difficult to solve dense models even with moderate size, and it is also the source of accumulating roundoff errors affecting the accuracy of the output. We present a new Sphere method, SM-6, for LP not using any pivot steps. The method is currently undergoing computational tests.

Downloads

Published

2012-12-12

How to Cite

Murty, K. G. (2012). Fast Descent Methods for LPs With No Matrix Inversions . Algorithmic Operations Research, 7(2). Retrieved from https://journals.lib.unb.ca/index.php/AOR/article/view/20394

Issue

Vol. 7 No. 2 (2012)

Section

Articles

License

The copyright of the submitted article is hereby transferred to “Preeminent Academic Facets Inc. (the “publisher”) to the extent possible by the laws in the respective countries of the author(s). The author(s) retain all other rights including the following: 1. Any patents and trademark rights. 2. Right to post a pre-print version of the paper on the author’s personal website which is not used for commercial purposes. The author(s) warrant that the work is original, it has not been published elsewhere, it is not currently under consideration for publication in any form, and it does not infringe any copyright laws. The signing author or agent further warrants that all authors, if applicable, have been informed of the terms of this agreement and obtained the approval and authority to sign the agreement.